Date: Fri, 25 Jan 2008 13:24:01 GMT From: John Baldwin <jhb@FreeBSD.org> To: Perforce Change Reviews <perforce@freebsd.org> Subject: PERFORCE change 134078 for review Message-ID: <200801251324.m0PDO1oc035171@repoman.freebsd.org>
next in thread | raw e-mail | index | archive | help
http://perforce.freebsd.org/chv.cgi?CH=134078 Change 134078 by jhb@jhb_zion on 2008/01/25 13:23:43 Don't set TDF_INTERRUPT to fake a KSE thread signal when a timeout triggers while the thread is still running. The KSE code controls setting and clearing TDF_INTERRUPT, so what would happen is that TDF_INTERRUPT would not be cleared until KSE finished sending a signal (i.e. never for non-KSE processes) but that while TDF_INTERRUPT is set, any sleeps that catch signals would immediately fail with EWOULDBLOCK. Thus, if a non-KSE process hit the race with a timeout firing then all future *sleep(... PCATCH ...) or cv_*wait_sig()'s would not actually block. Affected files ... .. //depot/projects/smpng/sys/kern/subr_sleepqueue.c#34 edit Differences ... ==== //depot/projects/smpng/sys/kern/subr_sleepqueue.c#34 (text+ko) ==== @@ -817,8 +817,7 @@ * sleepq_catch_signals(). */ if (TD_ON_SLEEPQ(td)) { - td->td_flags |= TDF_TIMEOUT | TDF_INTERRUPT; - td->td_intrval = EWOULDBLOCK; + td->td_flags |= TDF_TIMEOUT; thread_unlock(td); return; }
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200801251324.m0PDO1oc035171>